Merge pull request !3032 from DeshiChen/0711_akg_op_register_master

Since akg supports both Ascend and Gpu, but their supported type and format are different,
so we use two directory "ascend" and "gpu" to store their registers respectively, and use
an attribute "processor" to distinguish them.

Main changes:
1) Add two op register class "AkgAscendRegOp" and "AkgGpuRegOp", inherited from the original AkgRegOp.
2) Rewrite akg ascend op registers with new interface, move them into directory "ascend".
3) Rename the imply_type from "AutoDiff" to "AKG".
4) Modify function FindOp, check the processor when imply_type is "AKG".
5) Modify function CheckRepetition, remove the judgement for impl_path, check processor instead.

TODO: Remove op registers in akg root path.

Merge pull request !3035 from ZhangQinghua/master

Merge pull request !3027 from ZPaC/adaptation-for-ps-mode

Merge pull request !1898 from caifubi/data-dump

Merge pull request !3017 from ZPaC/add-ps-optimizer-kernel

Merge pull request !2925 from peixu_ren/custom_gpu

Merge pull request !2979 from ZPaC/add-push-pull-kernel

Merge pull request !2942 from lianliguang/convert-reduce-axis-when-selected-hw-special-format-for-reduce-kernel

Merge pull request !2962 from chenweifeng/smoothl1loss

Merge pull request !2964 from jiangzhenguang/Increase-the-max-tensor-size

Merge pull request !2971 from jjfeing/master

Merge pull request !2779 from chenweifeng/ctcloss

Merge pull request !2905 from kisnwang/parallel-reduce-sparse-gradient

Merge pull request !2862 from lianliguang/reselect-the-kernel-info-after-rectify-domask-kernel

Merge pull request !2835 from baihuawei/gatherV2

Merge pull request !2749 from VectorSL/argmaxwithvalue

Merge pull request !2804 from liuxiao93/ops-for-VM

1. add case to guard precision
2. fix a shape bug
3. fix a funcGraph bug

Merge pull request !2750 from TronZhang/clean_bot

Merge pull request !2701 from wangnan39/distinguish_cpu_sparse_operators_from_tbe